Every ageing system has three exits. Two of them are cheaper than people expect and one is far more expensive than the quote suggests. The difference between choosing well and choosing badly is a document nobody wants to pay for.
The routes, priced
| Phase or route | Cost | What it produces |
|---|---|---|
| Audit and scope | Rs 75,000 - Rs 3,00,000 | Inventory, risk list, rebuild plan |
| Data and integration move | Rs 1,50,000 - Rs 8,00,000 | Migrated records, tested connectors |
| Rebuild of the worst modules | Rs 6,00,000 - Rs 30,00,000 | Working replacements, old modules retired |
| Full assessment and options paper | Rs 8,00,000 - Rs 15,00,000 | Dependency map and a costed choice |
| Re-platform of the core | Rs 25,00,000 - Rs 80,00,000 | The same business, on a maintainable base |
| Parallel running and decommissioning | Rs 40,00,000 - Rs 1,50,00,000 | Dual operation, reconciliation, switch-off |
The last row is the one that kills budgets. Decommissioning is not a ceremony; it is proving the new system does everything the old one did, including the three reports nobody admits to relying on.
Patch, re-platform or rewrite
Patch means keep the system and fix the parts that hurt. It is the right answer more often than vendors admit, and it is the only route that can be finished in a quarter.
Re-platform keeps the business logic and replaces the foundation: database, runtime, hosting, authentication. It is the cheapest route that fixes unsupported technology, and it is usually invisible to users, which is both its strength and its political weakness.
Rewrite replaces the logic as well. It is the most expensive and the most likely to fail, not because the engineering is hard, but because the old system contains twenty years of decisions that nobody wrote down.
A rewrite that begins without an inventory is a rewrite that will rediscover each of those decisions in production.
Why the old system is expensive to leave
The cost is not the code. It is the knowledge.
Someone in the business knows that invoice numbering skips a series in March, that one customer’s credit limit is effectively double because of a 2014 agreement, and that the dispatch report must run before the stock report or the numbers disagree.
None of that is in a specification. All of it has to be found, written down and tested before the old system can be switched off. That discovery is what the audit is for, and it is why an audit at Rs 75,000 is the cheapest line in the table.
The reports nobody admits to using
Every old system has three or four outputs that exist nowhere in the documentation and are nonetheless load-bearing. A monthly statement that finance reconciles against the bank. A dispatch sheet that the warehouse prints at six in the morning. A stock valuation that only one person knows how to read.
They survive because they work, not because anyone designed them. When a replacement system produces a cleaner version, the complaint is not that it is wrong; it is that it is different, and the difference costs someone an hour a day.
Collect them during the audit, reproduce them exactly in the first release, and improve them later. Rebuilding the format in the same sprint as the migration is how a technically correct go-live gets rolled back.
The phasing that works
Replace one bounded area first, in parallel, and keep the old system running for everything else. Choose the area with the most pain and the fewest dependencies: usually a reporting module, a customer portal or a single workflow.
Run both for a month. Compare outputs daily. Only then retire the old piece.
Repeat. It is slower than a big-bang rewrite and it is roughly half the risk, because at every point there is a working system to fall back to.
What the options paper should contain
A list of every interface the system has, including the ones nobody remembers. A count of records by table. The three reports the business cannot operate without. An estimate for each of the three routes, with the assumptions stated.
If a vendor will not produce that document before quoting a rewrite, the quote is a guess with a payment schedule attached.
What the ceiling actually buys
At Rs 80,00,000 and above, the work stops being a software project and becomes an operations project: parallel running, staff retraining, data reconciliation over months, and a decommissioning plan that includes the auditors.
That is not padding. It is the difference between a system that goes live and a system that is trusted.
FAQ
Can we keep the old system and only replace the database? Yes, and it is a common first step. It fixes support and performance without changing behaviour, which makes it easy to justify.
Is a rewrite ever the right first move? When the platform is unsupported and the logic is simple. If both are true, a rewrite is defensible. If only one is, it usually is not.
How long does a phased replacement take? Nine to twenty-four months for a mid-sized system, with the first replacement live inside four.
What if we cannot find anyone who understands the old system? That is the strongest argument for the audit, and the audit will cost more because the knowledge has to be reconstructed from the data.
Do we need to stop the business during cutover? Not with parallel running. A weekend outage is normal for the final switch; a week is a planning failure.
Modernisation work in Noida
We run modernisation assessments from Noida and will recommend patching when that is the honest answer. Call 7840002466 or describe the system through the contact page.
The software cost comparison shows what a replacement build costs if that is where the assessment lands.
